AI agents & automation
Agents and bots that take over repetitive operational work, like EVA, our eligibility verification agent for U.S. healthcare.
WillCode Software
AI systems that
make it to production.
Eight years building LLM platforms, AI agents and backend systems for Motorola, Turing, Thoughtful and TOTVS. WillCode puts that experience to work on your product.
8+years shipping software
93%LLM response acceptance at Motorola, up from ~20%
5companies across Brazil and the U.S.
17professional certifications
Founder & CEO
Will Freitas
I'm William do Vale Cavalcante Freitas, a senior software engineer from Manaus, Brazil, specialized in backend and AI engineering. I started in test automation at INDT for Motorola, grew into deep learning and LLM work, and have since led and shipped AI systems for companies in Brazil and the U.S.
I founded WillCode Software to build AI systems with real business impact, both for clients and as our own products.
- Python
- FastAPI
- LLM systems
- Fine-tuning
- AWS · GCP
- Data pipelines
Services
What WillCode builds
LLM platforms
Multi-model orchestration, routing and cost control, fine-tuning, fast inference, and observability for prompts and responses.
Backend & data
Cloud-native APIs and high-throughput data pipelines in Python, FastAPI and PostgreSQL on AWS, GCP or Azure.
Products & projects
What we've built
Product
IronSystem
Training, diet and body-composition tracking where every session and meal earns XP. Climb ranks from E to S+, RPG-style.
ironsystem.fit
Client work · Thoughtful AI
EVA
Eligibility Verification Agent: an AI agent that automates insurance eligibility checks in U.S. healthcare workflows.
Built for Thoughtful AIExperience
Where I've worked
-
2025 – now
Backend AI Specialist · TOTVS
- Designed unified APIs that orchestrate multiple LLMs, with dynamic model selection, routing and cost optimization.
- Built monitoring and analytics pipelines for prompt/response tracking, performance and usage.
- Maintain cloud-native backend services for high-throughput, low-latency inference.
-
2025 – 2026
Forward Deployed Engineer · Thoughtful AI
- Designed and deployed automation bots for U.S. healthcare workflows, including EVA.
- Integrated OpenAI-powered decision-making and built Python data pipelines, deployed on AWS.
- Worked directly with clients to turn business requirements into AI solutions.
-
2025
Technical Lead · Turing
- Led the design of a large-scale data engineering and validation platform for training, fine-tuning and evaluating LLMs.
- Built high-throughput ingestion pipelines and AI-driven validation agents for data quality at scale.
-
2024
Freelance AI Software Engineer · Shock IT
- Fine-tuned a Donut DocVQA model to extract information from documents, served by a FastAPI backend on Azure.
-
2019 – 2025
INDT · for Motorola, 6 years
- Senior Software Engineer (2022–2025): deep learning models for Motorola's QA. Raised LLM response acceptance from ~20% to 93% by fine-tuning LLaMA models with Unsloth and optimizing inference with llama.cpp.
- Software Engineer II (2021–2022): led teams building algorithms that optimize test automation worldwide.
- Software Engineer I (2020–2021): built Motorola's worldwide mobile test automation tools.
- Test Automation Intern (2019–2020): Python, Selenium and Appium.
- Python
- FastAPI
- Django
- PostgreSQL
- PyTorch
- LLaMA
- Unsloth
- llama.cpp
- OpenAI
- AWS
- GCP
- Azure
- Docker
- Kubernetes
- Pandas
- RPA
Recognition
Education & awards
Education
B.Sc. Computer Engineering
Universidade do Estado do Amazonas (UEA)
AI Nanodegree, Udacity
Publication · SBC
Reusable Testing Actions on Test Case Automation for Mobile Devices
On the impact of code reuse in automated software testing.
Awards · INDT
Honorship for Excellent Work
2021, plus an Excellent Work award in 2020.
Contact
Have something to build?
Tell me about the problem you want solved.
williamdov.c.f@gmail.com